Sure -
Both Sun and Alpha planned to release next generation chips and systems in 4Q98. Neither did. SPARCIII has been plagued by delays and has been "readying for production" for 6 months now. Sooner or later they will make it.
This guy is obviously out of touch with Alpha development. He says Now compaq has not delivered any new version alpha at all ... Well, the EV6 version of the chip was released to production 3 months ago and the "speed-enhanced" 800MHz version is scheduled for this summer. I think both of those are "new versions". Alpha's record on meeting silicon schedule is a lot better than Sun's.
As far as scalability goes, I assume that he is talking about the UE10000. Well, this was not developed by Sun - they bought the technology from Cray. The UE10000 is not compatible with any other Sun machines even at the OS level, and OS support has been weak. Sun has done a great job of selling the "sizzle" on the UE10000 but actual machine sales have been in the low hundreds.
The CPQ offering for a large scalable machine is a thing called "wildfire", also supposed to be a 4Q98 product. There has been little public discussion about this machine so I know little about the architecture except that it is a 32- or 64-way SMP in its first incarnation and later will go to 256 way.
The history of Alpha has been that they offer about twice the performance of SPARC for the same cost, and I have seen nothing that says that will change any time in the near future. In addition, Alpha machines also run NT, which is something that will never happen on SPARC.
Either this guy is genuinely uninformed about what else is going on in the industry (often true about SUN guys) or he is intentionally misrepresenting the facts. In any event CPQ will continue to trump Sun on performance, and Sun will continue to enjoy the almost fanatical Unix base they have established over the years. It is also true that no one but Sun uses SPARC...
He is correct on SPARC advantages over Merced and even McKinley, which will not offer comparable capability for many years to come. Alpha has had a similar advantage for more than 5 years and Intel is still the microprocessor the fans are screaming for. Most Intel users will not care - they want a 64 bit machine which runs all their MSFT software and could care less about SPARC.
Alpha may have a play because of Intel's delay of their 64 bit parts, since Alpha will be the development platform for 64 bit NT, and not Intel. Maybe CPQ will be able to take advantage of that, either with 64 bit Microsoft products on Alpha, or by grabbing the "64 bit" mantle in that space in a more general way. No other NT vendor has a shot since they have no chips. There was a lot of discussion about that possibility last fall but I have not heard much recently. |
